This is patently untrue. Choosing the right silicon for a new project is an immense task. Selecting the micro for even a small job (except in special cases) requires me to generate a mountain of paperwork. I have to show a matrix of choices, pro/con points, cost targets, how that will affect the BOM of the board, expected EMI issues, first-order approximation of power budget, etc.
A vendor who puts me through 20 minutes of registration (especially on Xilinx's bouncy servers - here one second, down the next - and braindead browser requirements "you must be running IE6 from Win XP SP2 and you must have the patches up to 9/1/2005 installed BUT NOTHING LATER, and you must have all security disabled and some unstated ActiveX control installed") is going to be kicked into the "too hard" pile immediately.
Special case exceptions I referred to above are projects that will reuse a lot of code from an existing device, so the micro choice is constrained.